# Framework Conventions Function as Executable Context for Agents ## Claim An opinionated framework's conventions narrow a coding agent's valid choices the same way documentation would, which is why agents work better inside batteries-included frameworks than inside unopinionated ones. <!-- atomicity: 1 idea ✓ --> ## Scope & conditions - Holds for frameworks with established, discoverable conventions (known directory structure, standard services, documented patterns) that an agent can pattern-match against. - Breaks for greenfield or highly custom architectures with no established convention to follow — there, an agent has nothing to narrow its choices against, opinionated framework or not. - Does not establish the size of the advantage — no cross-framework benchmark exists yet; this is a mechanism claim, not a measured effect size. ## Evidence Taylor Otwell and Matt Stauffer describe Laravel's built-in database access, standard directory structure, and documented patterns as reducing the number of architectural decisions an agent has to make before it writes useful code. Laravel Boost extends this by feeding agents live framework documentation and project context directly, rather than leaving them dependent on what the underlying model happened to memorize during training — addressing the separate but related problem of stale model knowledge. ## Links - [[3. Resources/Permanent Notes/Ship an Opinionated Default Expose One Knob]] — because extends; this claim is the agent-specific instance of that general design principle - [[3. Resources/Permanent Notes/Context Not Constraints Is What Unhobbles an Agent]] — because evidence-for; fresh, accurate context (not just narrower constraints) is what Boost supplies - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] — because applies-in; framework and tooling design choices that shape agent-assisted development outcomes ## Tests / falsifiers - Run the same coding task through an agent in Laravel vs. an unopinionated framework with equivalent documentation quality, and compare first-pass correctness. If they converge, the conventions themselves aren't doing the work — documentation quality is. - If agents perform equally well in opinionated frameworks with stale docs vs. current docs (Boost-fed), the "fresh context" half of this claim would be the weaker half, not the conventions half. ## Maintenance - Last tended: 2026-08-16 - Next review: 2026-11-16 ## Source - [[3. Resources/Readwise/Articles/Taylor Otwell on Why Laravel + AI Pair So Well, and the Future of Programming]] — YouTube interview, Matt Stauffer and Taylor Otwell.